Thoughts on CTC's Latest CPMr Report

· Articles

Thoughts on CTC's Latest CPMr Report

Phil here, Director of Paid Social. Common Thread Collective publishes some really great work within DTC. Their writing has raised the floor for all of us.

That track record is exactly why thisparticular report deserves a careful second look.

The headline and the data do not agree.

We want to walk through why we read it differently.

Where We Agree

CTC is right that CPMr needs more scrutiny in DTC conversations. They are right that attributed ROAS is a poor proxy for incremental value. They are right that the 1.2x default factor most agencies apply is lazy and that custom lift-tested factors are the only honest way to talk about incrementality.

Every one of those points is a real contribution.

The report sparking conversation is a net positive. Our concern is not that it exists. It is that the specific conclusion being shared, that higher CPMr delivers better incremental ROAS, is not what the data in the report actually shows.

The Core Issue

At the campaign level, iROAS is just attributed ROAS multiplied by the brand's incrementality factor. Every campaign inside a brand gets the same multiplier. That means the rank order of campaigns by iROAS is mathematically identical to the rank order by attributed ROAS.

The report concedes this directly in Section 08: "iROAS conclusions are identical to ROAS (same ranks)."

So when the within-brand analysis concludes that "higher CPMr delivers better iROAS," what it actually shows is that higher CPMr campaigns have higher attributed ROAS.

The incrementality adjustment is doing no work at the campaign level.

The headline claim and the methodology point in different directions.

What the Significant Finding Actually Says

There is one statistically significant result in the report.

It is the cross-brand analysis in Section 05.

Low-CPMr brands average a 1.36x incrementality factor.

High-CPMr brands average 1.01x. Spearman ρ = −0.62, p = 0.009.

That is a 35% gap in incremental efficiency favoring lower CPMr. It is the only result in the entire report that clears a meaningful significance threshold. And it points the opposite direction from the headline.

The within-brand finding that carries the headline runs at ρ = +0.18, Wilcoxon p = 0.35.

The report calls this "directionally consistent."

At n=17, a 12-5 directional split happens roughly 7% of the time by chance alone.

This is not a finding you can build a DTC narrative on.

The Other Data Points That Deserve Attention

Quartile dose-response is flat:

Q1 iROAS 2.07x. Q2 2.18x. Q3 2.00x. Q4 2.19x.

Q3 sits below Q1.

The gap from lowest to highest quartile is 0.12x. There is no dose-response curve here.

The quartile results table is missing data:

The Avg CPMr and Eff. Factor columns are filled with em-dashes.

The sensitivity analysis collapses the effect:

At Spend ≥ $2.5K: 47% win rate for high CPMr.

At Spend ≥ $5K: 46%. At credible spend thresholds, it is a coin flip.

The headline only holds when small low-spend campaigns are included in the sample.

Frequency is the unaddressed mechanism:

Within-brand CPMr and Frequency correlate at ρ = +0.86.

They are effectively the same signal inside a brand.

Nothing in the analysis controls for frequency, so the claim reduces to "higher frequency campaigns have higher attributed ROAS," which is a well-documented attribution inflation pattern, not an incrementality finding.

Retargeting is not separated from prospecting:

Retargeting campaigns carry high frequency, high CPMr, and high attributed ROAS, while delivering the lowest true incrementality of any campaign type.

Pooling all ACQ campaigns under an account-level factor cannot surface this.

Potential conclusion: The "high CPMr wins" pattern is most likely a "retargeting wins" pattern.

Reverse causation is not tested:

Campaigns that perform get more budget.

More budget drives frequency up, which drives CPMr up.

The causal arrow likely runs from performance to CPMr, not CPMr to performance.

The report does not address this.

Blended iROAS conflates customer recency with incrementality:

New vs returning is a different axis than incremental vs non-incremental.

Treating them as the same weakens the framework.

Sample sizes are small:

7 brands in the headline analysis. 8 brands in the quartile analysis.

Conclusions are being drawn at significance levels (p = 0.35) that would not survive peer review in any quantitative field.

Why This Matters

The report is circulating with the headline "higher CPMr delivers better incremental ROAS" as its takeaway. Brands and agencies reading it at the surface level will conclude that frequency saturation is not a problem, that rising CPMr is a healthy signal, and that scaling without regard to reach efficiency is defensible.

That is a conclusion the data does not support.

We have just finished a Q1 2026 audit of our own book covering 7 DTC verticals. 30.5% of Q1 spend across those verticals went to impressions beyond the 2nd view per person.

In high-frequency accounts, estimated incremental CPA runs 2 to 3x reported CPA. These are patterns CTC's own cross-brand finding (ρ = −0.62) is consistent with.

The data does not support the "keep scaling, ignore CPMr" read.

Our Honest Version of the Finding

If we had to write a headline from CTC's own data, it would be closer to this:

"Brands that reach audiences cheaply drive more incremental revenue at the account level. 

Within a brand, campaigns with higher attributed ROAS also tend to have higher CPMr, but this is an attribution artifact driven by frequency, not evidence that higher CPMr causes better incrementality."

That is a more accurate summary.

It is also a less shareable one.

Why We Are Writing This

CTC is good at this work. That is why we think this specific piece deserves a closer read before it becomes settled DTC wisdom.

Data should not travel faster than the caveats attached to it.

We would rather have the conversation than avoid it.

Getting the framework right matters more than any single report, ours included.

Phil and team, Taikun Digital


Back to all articles